A new framework for landslide susceptibility mapping in contiguous impoverished areas using machine learning and...

Area of Science:

• Geosciences
• Environmental Science
• Disaster Management

Background:

• Landslides pose significant global risks, necessitating accurate landslide susceptibility mapping (LSM) for effective disaster prevention.
• Impoverished regions like Liangshan, Sichuan, are particularly vulnerable to geological disasters, requiring tailored risk assessment strategies.

Purpose of the Study:

• To develop and evaluate a novel LSM framework integrating random forest (RF), support vector machine (SVM), and catastrophe theory (CT).
• To assess landslide susceptibility in the impoverished areas of Liangshan, Sichuan, using the proposed integrated models.
• To enhance the accuracy and reliability of LSM for improved disaster prevention and sustainable development.

Main Methods:

• Selection of 12 factors representing internal environmental and external triggering conditions for landslide susceptibility.
• Application of the frequency ratio method to correlate historical landslides with selected factors.
• Integration of catastrophe theory (CT) with RF and SVM models to create RF-CT and SVM-CT frameworks for LSM.
• Evaluation of model performance using the receiver operating characteristic (ROC) curve analysis.

Main Results:

• The integrated RF-CT and SVM-CT models demonstrated superior performance compared to individual RF and SVM models.
• RF-CT achieved a success rate of 0.899 and a prediction rate of 0.783.
• SVM-CT achieved a success rate of 0.873 and a prediction rate of 0.775, showing notable improvements.
• Both integrated models showed approximately 10% improvement in success rate and 5% in prediction rate.

Conclusions:

• The novel RF-CT and SVM-CT frameworks provide a more accurate and reliable approach to landslide susceptibility mapping.
• The findings offer valuable data for disaster prevention, poverty alleviation, and sustainable development planning in Liangshan, Sichuan.
• This integrated approach can be adapted for landslide risk assessment in similar vulnerable regions globally.